Chemical reactions up close

chemical reactions

Editor's Introduction

Direct imaging of covalent bond structure in single-molecule chemical reactions

annotated by

How exactly do chemicals react? How do molecules break and form new bonds? How do we know what the products of chemical reactions look like? In this study, the organic hydrocarbon, phenylene-1,2-ethynylene (C26H14) is visualized at the atomic level with non-contact atomic force microscopy (nc-AFM).  The reactant, phenylene-1,2-ethynylene (C26H14), was placed on a solid silver surface and heated, causing the products to form. The scientists saw a few products being formed from the reactant and they were able to distinguish the structure and bond order of these products using nc-AFM. To determine how exactly these three products were formed, computational chemistry was used to examine how the reactant was converted to the products.

Paper Details

Original title
Direct imaging of covalent bond structure in single-molecule chemical reactions
Original publication date
Vol. 340 no. 6139 pp. 1434-1437
Issue name


Observing the intricate chemical transformation of an individual molecule as it undergoes a complex reaction is a long-standing challenge in molecular imaging. Advances in scanning probe microscopy now provide the tools to visualize not only the frontier orbitals of chemical reaction partners and products, but their internal covalent bond configurations as well. We used noncontact atomic force microscopy to investigate reaction-induced changes in the detailed internal bond structure of individual oligo-(phenylene-1,2-ethynylenes) on a (100) oriented silver surface as they underwent a series of cyclization processes. Our images reveal the complex surface reaction mechanisms underlying thermally induced cyclization cascades of enediynes. Calculations using ab initio density functional theory provide additional support for the proposed reaction pathways.


Understanding the microscopic rearrangements of matter that occur during chemical reactions is of great importance for catalytic mechanisms and may lead to greater efficiencies in industrially relevant processes (12). However, traditional chemical structure characterization methods are typically limited to ensemble techniques where different molecular structures, if present, are convolved in each measurement (3). This limitation complicates the determination of final chemical products, and often it renders such identification impossible for products present only in small amounts. Single-molecule characterization techniques such as scanning tunneling microscopy (STM) (45) potentially provide a means for surpassing these limitations. Structural identification using STM, however, is limited by the microscopic contrast arising from the electronic local density of states (LDOS), which is not always easily related to chemical structure. Another important subnanometer-resolved technique is transmission electron microscopy (TEM). Here, however, the high-energy electron beam is often too destructive for organic molecule imaging. Recent advances in tuning fork–based noncontact atomic force microscopy (nc-AFM) provide a method capable of nondestructive subnanometer spatial resolution (610). Single-molecule images obtained with this technique are reminiscent of wire-frame chemical structures and even allow differences in chemical bond order to be identified (10). Here we show that it is possible to resolve with nc-AFM the intramolecular structural changes and bond rearrangements associated with complex surface-supported cyclization cascades, thereby revealing the microscopic processes involved in chemical reaction pathways.

Intramolecular structural characterization was performed on the products of a thermally induced enediyne cyclization of 1,2-bis((2-ethynylphenyl)ethynyl)benzene (1). Enediynes exhibit a variety of radical cyclization processes known to compete with traditional Bergman cyclizations (1112), thus often rendering numerous products with complex structures that are difficult to characterize using ensemble techniques (13). To directly image these products with subnanometer spatial resolution, we thermally activated the cyclization reaction on an atomically clean metallic surface under ultrahigh vacuum (UHV). We used STM and nc-AFM to probe both the reactant and final products at the single-molecule level. Our images reveal how the thermally induced complex bond rearrangement of 1 resulted in a variety of unexpected products, from which we have obtained a detailed mechanistic picture—corroborated by ab initio density functional theory (DFT) calculations—of the cyclization processes. 

We synthesized 1,2-bis((2-ethynylphenyl)ethynyl)benzene (1) through iterative Sonogashira cross-coupling reactions (scheme S1) (14). We deposited 1 from a Knudsen cell onto a Ag(100) surface held at room temperature under UHV. Molecule-decorated samples were transferred to a cryogenic imaging stage (T ≤ 7 K) before and after undergoing a thermal annealing step. Cryogenic imaging was performed both in a home-built T = 7 K scanning tunneling microscope and in a qPlus-equipped (1516) commercial Omicron LT-STM/AFM at T = 4 K. nc-AFM images were recorded by measuring the frequency shift of the qPlus resonator while scanning over the sample surface in constant-height mode. For nc-AFM measurements, the apex of the tip was first functionalized with a single CO molecule (6). To assess the reaction pathway energetics, we performed ab initio DFT calculations within the local density approximation (17) using the GPAW (Grid-based Projector Augmented Wavefunction) code (1819).

Figure 1A shows a representative STM image of 1 on Ag(100) before undergoing thermal annealing. The adsorbed molecules each exhibited three maxima in their LDOS at positions suggestive of the phenyl rings in 1(Fig. 1A). Annealing the molecule-decorated Ag surface up to 80°C left the structure of the molecules unchanged. Annealing the sample at T ≤ 90°C, however, induced a chemical transformation of 1 into distinctively different molecular products (some molecular desorption was observed)Figure 1B shows an STM image of the surface after annealing at 145°C for 1 min. Two of the reaction products can be seen in this image, labeled as 2and 3. The structures of the products are unambiguously distinguishable from one another and from the starting material 1, as shown in the close-up STM images of the most common products 2, 3, and 4 in Fig. 2, B to D. The observed product ratios are 2:3:4 = (51 ± 7%):(28 ± 5%):(7 ± 3%), with the remaining products comprising other minority monomers as well as fused oligomers (fig. S2) (14).


Fig 1. STM images of a reactant-decorated Ag(100) surface before and after thermally induced cyclization reactions.  (A) Constant-current STM image of 1 as deposited on Ag(100) (I= 25 pA, V = 0.1 V, T = 7 K). A model of the molecular structure is overlaid on a close-up STM image. (B) STM image of products 2 and 3 on the surface shown in (A) after annealing at T = 145°C for 1 min (I = 45 pA, V = 0.1 V, T = 7 K).

The researchers wanted to see if products would form after heating their reactant on the silver surface. Researchers also wanted to know what the products would form, how many there would be, and what the chemical structure would be.


Panel A shows the STM image of the reactant molecules. The reaction occurs on a silver plate, which allows the molecules to be viewed using STM.

Panel B shows what the molecules look like after the sample was heated at temperatures over 90ºC after heating this sample at 145ºC for 1 minute. They found a mixture of products being formed. We can see from these images that product 2 and 3 have different shapes

Nobel Prize for the STM

Virtual Lab to Explore how STM works


These images show that their method is working; the scientists can tell the differences from the products and reactants using STM and tell that they have multiple products forming from this reaction.


Fig. 2.  Comparison of STM images, nc-AFM images, and structures for molecular reactant and products. (A) STM image of 1 on Ag(100) before annealing. (B to D) STM images of individual products 2, 3, and 4 on Ag(100) after annealing at T > 90°C (I = 10 pA, V = –0.2 V, T = 4 K). (E) nc-AFM image of the same molecule (reactant 1) depicted in (A). (F to H) nc-AFM images of the same molecules (products 2, 3, and 4) depicted in (B) to (D). nc-AFM images were obtained at sample bias V = –0.2 V (qPlus sensor resonance frequency = 29.73 kHz, nominal spring constant = 1800 N/m, Q-value = 90,000, oscillation amplitude = 60 pm). (I to L) Schematic representation of the molecular structure of reactant 1 and products 2, 3, and 4. All images were acquired with a CO-modified tip.


Researchers want to know which technique—STM or nc-AFM—would be suitable to see the reactant, as well as any products that may form after heating the reactant on the silver surface. 


The first two rows of images show the results of using the two microscopy techniques, STM and nc-AFM.

Panels A, E, and I show the reactant;

panels B, F, and J show product 2;

panels C, G, and K show product 3;

and panels D, H, and L show product 4.

The top row shows the reactant and 3 products that form, visualized with STM. The STM images are quite blurry; it is difficult to see the bonds.

The second row shows the AFM images. Now we can see the bonds and shapes much more clearly.

The final row shows the textbook line-angle formula drawings of the reactant and products that closely resemble the nc-AFM images.

A video demonstrating how nc_AFM works

Virtual Lab to Explore how AFM works


nc-AFM can be used to view these molecules. The geometry of the products can be determined using this technique. We can see that it offers greater resolution than STM.

Detailed subnanometer-resolved structure and bond conformations of the molecular reactant 1 and products (2,3, and 4) were obtained by performing nc-AFM measurements of the molecule-decorated sample both before and after annealing at T ≥ 90°C. Figure 2E shows a nc-AFM image of 1 before annealing. In contrast to the STM image (Fig. 2A) of 1, which reflects the diffuse electronic LDOS of the molecule, the AFM image reveals the highly spatially resolved internal bond structure. A dark halo observed along the periphery of the molecule is associated with long-range electrostatic and van der Waals interactions (620). The detailed intramolecular contrast arises from short-ranged Pauli repulsion, which is maximized in the regions of highest electron density (20). These regions include the atomic positions and the covalent bonds. Even subtle differences in the electron density attributed to specific bond orders can be distinguished (10), as evidenced by the enhanced contrast at the positions of the triple bonds within 1. This effect is to be distinguished from the enhanced contrast observed along the periphery of the molecule, where spurious effects (e.g., because of a smaller van der Waals background, enhanced electron density at the boundaries of the delocalized π-electron system, and molecular deviations from planarity) generally influence the contrast (1020).

Figure 2, F to H, shows subnanometer-resolved nc-AFM images of reaction products 2, 3, and 4 that were observed after annealing the sample at T > 90°C. The structure of these products remained unaltered even after further annealing to temperatures within the probed range from 90° to 150°C (150°C was the maximum annealing temperature explored in this study). The nc-AFM images reveal structural patterns of annulated six-, five-, and four-membered rings. The inferred molecular structures are shown in Fig. 2, J to L. Internal bond lengths measured by nc-AFM have previously been shown to correlate with Pauling bond order (10), but with deviations occurring near a molecule’s periphery, as described above. As a result, we could extract clear bonding geometries for the products (Fig. 2, J to L), but not their detailed bond order. The subtle radial streaking extending from the peripheral carbon atoms suggests that the valences of the carbon atoms are terminated by hydrogen (20). This is in agreement with molecular mass conservation for all proposed product structures and indicates that the chemical reactions leading to products 2, 3, and 4 are exclusively isomerization processes (21).

Our ability to directly visualize the bond geometry of the reaction products (Fig. 2, F to H) provides insight into the detailed thermal reaction mechanisms that convert 1 into the products. We limit our discussion to the reaction pathways leading from 1 to the two most abundant products, 2 and 3. The reactivity of oligo-1,2-diethynylbenzene 1 can be rationalized by treating the 1,2-diethynylbenzene subunits as independent but overlapping enediyne systems that are either substituted by two phenyl rings for the central enediyne, or by one phenyl ring and one hydrogen atom in the terminal segments. This treatment suggests three potential cyclizations along the reaction pathway (resulting in six-, five-, or four-membered rings) (22), in addition to other possible isomerization processes such as [1,2]-radical shifts or bond rotations that have been observed in related systems (23). Combinations of these processes leading to the products in a minimal number of steps were explored and analyzed using DFT calculations (14).

We started by calculating the total energy of a single adsorbed molecule of the reactant 1 on a Ag(100) surface. Activation barriers and the energy of metastable intermediates were calculated (including molecule-surface interactions) for a variety of isomeric structures along the reaction pathway leading toward the products 2 and 3. Our observations that the structure of reactants on Ag(100) remains unchanged for T < 90°C and that no reaction intermediates can be detected among the products indicate that the initial enediyne cyclization is associated with a notable activation barrier that represents the rate-determining step in the reaction. In agreement with experiments, DFT calculations predict an initial high barrier for the first cyclization reactions, followed by a series of lower barriers associated with subsequent bond rotations and hydrogen shifts.

Figure 3A shows the reaction pathway determined for the transformation of 1 into product 2. The rate-determining activation barrier is associated with a C1–C6 Bergman cyclization of a terminal enediyne coupled with a C1–C5 cyclization of the internal enediyne segment to give the intermediate diradical Int1 in an overall exothermic process (–60.8 kcal mol–1). Rotation of the third enediyne subunit around a double bond, followed by the C1–C5 cyclization of the fulvene radical with the remaining triple bond, leads to Int2. The rotation around the exocyclic double bond is hindered by the Ag surface and requires the breaking of the bond between the unsaturated valence on the sp2 carbon atom and the Ag. Yet the activation barrier associated with this process does not exceed the energy of the starting material used as a reference. The formation of three new carbon-carbon bonds and the extended aromatic conjugation stabilize Int2 by –123.9 kcal mol–1 relative to 1. Finally, a sequence of radical [1,2]- and [1,3]-hydrogen shifts followed by a C1–C6 cyclization leads from Int2 directly to the dibenzofulvalene 2. Our calculations indicate that the substantial activation barriers generally associated with radical hydrogen shifts in the gas phase (50 to 60 kcal mol–1) (24, 25) are lowered through the stabilizing effect of the Ag atoms on the surface, and thus they do not represent a rate-limiting process (Fig. 3A).


Fig. 3.  Reaction pathways and their associated energies as calculated by DFT. (A) Proposed pathway for the cyclization of reactant 1 into product 2 on Ag(100). (B) Proposed pathway for the cyclization of reactant 1 into product 3 on Ag(100). Energies for 1, 2, and 3(end point solid circles), for the intermediates Int1, Int2, Int3,tInt1, and tInt2 (intermediate solid circles), and for the reaction barriers (open circles) were calculated using ab initio DFT theory (14). Ball-and-stick models show the nonplanar structure of intermediates. The symbol ‡ indicates the rate-determining transition state; the red line is the reference energy of 1 on Ag(100).


What reaction pathways were being used to convert the reactant into two of the products?

What sorts of intermediates were being formed?

Results for Panel A

The energy pathways associated with making product 2. The intermediates of the pathway are shown. These pathways were calculated with computational chemistry methods. The single-headed arrow shows how one electron moves from atom to atom. The rate-determining step is the conversion of the reactant to the first intermediate molecule, the cyclization reaction. This step is an exothermic process (-60.8 kcal mol-1). 

Results for Panel B

This shows the energy pathway associated with making product 3. The rate-determining step in this reaction is also the conversion of the reactant to the first intermediate molecule. Product 3 is formed due to the silver atoms on the surface, which help to stabilize the reaction by causing the molecule to arrange in a non-planar shape, allowing hydrogen to shift and cyclization reactions to occur.

Nobel Prize for DFT, the computational chemistry technique used:


Overall, for each product, the activation barriers associated with each intermediate are smaller than the activation barriers for the formation of the first intermediate. The large amount of energy needed for the initial reaction is lowered due to silver on the surface helping to stabilize the reaction. The formation of 2 and 3 follow a C1-C5 enediyne cyclization, which is less stable than the C1-C6 cyclization. The steric bulk from the substituents around the triple bonds and the presence of silver allow for the formation of the C1-C5 cyclization products.

The reaction sequence toward 3 is illustrated in Fig. 3B. The rate-determining first step involves two C1–C5cyclizations of the sterically less hindered terminal enediynes to yield benzofulvene diradicals. The radicals localized on the exocyclic double bonds subsequently recombine in a formal C1–C4 cyclization to yield the four-membered ring in the transient intermediate tInt1. This process involves the formation of three new carbon-carbon bonds, yet it lacks the aromatic stabilization associated with the formation of the naphthyl fragment inInt2 and is consequently less exothermic (–60.7 kcal mol–1). A sequence of bond rotations transforms tInt1 viaInt3 into tInt2. Alignment of the unsaturated carbon valences in diradical tInt1 with underlying Ag atoms maximizes the interaction with the substrate and induces a highly nonplanar arrangement, thereby making subsequent rotations essentially barrierless. [1,2]-Hydrogen shifts and a formal C1–C6 cyclization yield the biphenylene 3.

Both reaction pathways toward 2 and 3 involve C1–C5 enediyne cyclizations. These are generally energetically less favorable relative to the preferred C1–C6 Bergman cyclizations (22), but factors such as the steric congestion induced by substituents on the alkynes (1222), the presence of metal catalysts (26), or single-electron reductions of enediynes (2728) have been shown to sway the balance toward C1–C5 cyclizations yielding benzofulvene diradicals. All three of these factors apply to the present case of the thermally induced cyclization of 1 on Ag(100) [e.g., bulky phenyl substituent on C1 and C6, a metallic substrate, and a charge transfer of 0.5 electrons from the substrate to 1 (14)], thus facilitating the C1–C5 cyclizations. The precise order of the low-energy processes (such as the Int3-tInt2 rotation and the tInt2-3 [1,2]-hydrogen shifts) following the rate-limiting initial cyclizations cannot be strictly determined experimentally. However, the sequence does not change the overall reaction kinetics and thermodynamics discussed above. Our bond-resolved single-molecule imaging thus allows us to extract an exhaustive picture and unparalleled insight into the chemistry involved in complex enediyne cyclization cascades on Ag(100) surfaces. This detailed mechanistic understanding in turn guides the design of precursors for the rational synthesis of functional surface-supported molecular architectures.

References and Notes

  1. G. Ertl, Elementary steps in heterogeneous catalysis. Angew. Chem. Int. Ed. Engl. 29, 1219 (1990).

  2. G. A. Somorjai, The surface science of heterogeneous catalysis. Surf. Sci. 299–300, 849 (1994).

  3. D. A. Skoog, F. J. Holler, S. R. Crouch, Principles of Instrumental Analysis (Brooks/Cole, Belmont, CA, 2006).

  4. R. Wiesendanger, Scanning Probe Microscopy and Spectroscopy: Methods and Applications (Cambridge Univ. Press, Cambridge, 1998).

  5. S.-W. Hla, L. Bartels, G. Meyer, K.-H. Rieder, Inducing all steps of a chemical reaction with the scanning tunneling microscope tip: Towards single molecule engineering. Phys. Rev. Lett. 85, 2777 (2000).

  6. L. Gross, F. Mohn, N. Moll, P. Liljeroth, G. Meyer, The chemical structure of a molecule resolved by atomic force microscopy. Science 325, 1110 (2009).

  7. L. Gross et al., Organic structure determination using atomic-resolution scanning probe microscopy. Nat. Chem. 2, 821 (2010).

  8. K. Ø. Hanssen et al., A combined atomic force microscopy and computational approach for the structural elucidation of breitfussin A and B: Highly modified halogenated dipeptides from Thuiaria breitfussi. Angew. Chem. Int. Ed. 51, 12238 (2012).

  9. N. Pavliček et al., Atomic force microscopy reveals bistable configurations of dibenzo[a,h]thianthrene and their interconversion pathway. Phys. Rev. Lett. 108, 086101 (2012).

  10. L. Gross et al., Bond-order discrimination by atomic force microscopy. Science 337, 1326 (2012).

  11. R. R. Jones, R. G. Bergman, p-Benzyne. Generation as an intermediate in a thermal isomerization reaction and trapping evidence for the 1,4-benzenediyl structure. J. Am. Chem. Soc. 94, 660 (1972).

  12. C. Vavilala, N. Byrne, C. M. Kraml, D. M. Ho, R. A. Pascal Jr., Thermal C1-C5 diradical cyclization of enediynes. J. Am. Chem. Soc. 130, 13549 (2008).

  13. J. P. Johnson et al., Comparison of “polynaphthalenes” prepared by two mechanistically distinct routes. J. Am. Chem. Soc. 125, 14708 (2003).

  14. See supplementary materials on Science Online.

  15. F. J. Giessibl, Advances in atomic force microscopy. Rev. Mod. Phys. 75, 949 (2003).

  16. F. J. Giessibl, Atomic resolution on Si(111)-(7×7) by noncontact atomic force microscopy with a force sensor based on a quartz tuning fork. Appl. Phys. Lett. 76, 1470 (2000).

  17. J. P. Perdew, A. Zunger, Self-interaction correction to density-functional approximations for many-electron systems. Phys. Rev. B 23, 5048 (1981).

  18. J. J. Mortensen, L. B. Hansen, K. W. Jacobsen, Real-space grid implementation of the projector augmented wave method. Phys. Rev. B 71, 035109 (2005).

  19. J. Enkovaara et al., Electronic structure calculations with GPAW: A real-space implementation of the projector augmented-wave method. J. Phys. Condens. Matter 22, 253202 (2010).

  20. N. Moll, L. Gross, F. Mohn, A. Curioni, G. Meyer, The mechanisms underlying the enhanced resolution of atomic force microscopy with functionalized tips. New J. Phys. 12, 125020 (2010).

  21. Figure 2G suggests a four-membered ring between two six-membered rings, but does not resolve it perfectly. DFT calculations indicate that other structural isomers, such as an eight-membered ring next to a six-membered ring, are energetically very unfavorable relative to the structure of 3.

  22. M. Prall, A. Wittkopp, P. R. Schreiner, Can fulvenes form from enediynes? A systematic high-level computational study on parent and benzannelated enediyne and enyne-allene cyclizations. J. Phys. Chem. A 105, 9265 (2001).

  23. K. D. Lewis, A. J. Matzger, Bergman cyclization of sterically hindered substrates and observation of phenyl-shifted products. J. Am. Chem. Soc. 127, 9968 (2005).

  24. M. A. Brooks, L. T. Scott, 1,2-Shifts of hydrogen atoms in aryl radicals. J. Am. Chem. Soc. 121, 5444 (1999).

  25. M. Hofmann, H. F. Schaefer, Pathways for the reaction of the butadiene radical cation, [C4H6]•+, with ethylene. J. Phys. Chem. A 103, 8895 (1999).

  26. C.-Y. Lee, M.-J. Wu, Synthesis of benzofulvenes by palladium-catalyzed cyclization of 1,2-dialkynylbenzenes. Eur. J. Org. Chem. 2007, 3463 (2007).

  27. V. Alabugin, M. Manoharan, Radical-anionic cyclizations of enediynes: Remarkable effects of benzannelation and remote substituents on cyclorearomatization reactions. J. Am. Chem. Soc. 125, 4495 (2003).

  28. V. Alabugin, S. V. Kovalenko, C1-C5 photochemical cyclization of enediynes. J. Am. Chem. Soc. 124, 9052 (2002).

  29. R. H. Grubbs, D. Kratz, Highly unsaturated oligomeric hydrocarbons: α-(Phenylethynyl)-ω-phenylpoly[1,2-phenylene(2,1-ethynediyl)]. Chem. Ber. 126, 149 (1993).

  30. Horcas et al., WSXM: A software for scanning probe microscopy and a tool for nanotechnology. Rev. Sci. Instrum. 78, 013705 (2007).

  31. Zhao et al., Controlling the Kondo effect of an adsorbed magnetic ion through its chemical bonding. Science 309, 1542 (2005).

  32. W. Kohn, Nobel Lecture: Electronic structure of matter—wave functions and density functionals. Rev. Mod. Phys. 71, 1253 (1999).

  33. R. F. W. Bader, Atoms in Molecules: A Quantum Theory (Oxford Univ. Press, Oxford, 1990), vol. 3.

  34. J. L. Cabellos et al., Understanding charge transfer in donor–acceptor/metal systems: A combined theoretical and experimental study. J. Phys. Chem. C 116, 17991 (2012).

  35. Acknowledgments: Supported by the Office of Naval Research BRC Program (molecular synthesis, characterization, and STM imaging); the Helios Solar Energy Research Center supported by the Office of Science, Office of Basic Energy Sciences, U.S. Department of Energy under contract DE-AC02-05CH11231 (STM and nc-AFM instrumentation development, AFM operation); NSF grant DMR-1206512 (image analysis); and European Research Council advanced grant DYNamo ERC-2010-AdG-267374 (ab initio calculations). Computing time was provided by the Barcelona Supercomputing Center “Red Española de Supercomputacion.” D.G.d.O. acknowledges fellowship support by the European Union under FP7-PEOPLE-2010-IOF-271909, A.R. by Austrian Science Fund (FWF) grant J3026-N16, and D.J.M. by the Spanish “Juan de la Cierva” program (JCI-2010-08156). The data presented in the manuscript are tabulated in the main paper and in the supplementary materials. The authors declare no conflicts of interest.